Design Challenges in Low-resource Cross-lingual Entity Linking

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 6418–6432, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 6418 Design Challenges in Low-resource Cross-lingual Entity Linking Xingyu Fu1∗, Weijia Shi2∗, Xiaodong Yu1, Zian Zhao3, Dan Roth1 1University of Pennsylvania 2University of California Los Angeles 3Columbia University {xingyuf2, xdyu, danroth}@seas.upenn.edu swj0419@g.ucla.edu, zz2562@columbia.edu Abstract Cross-lingual Entity Linking (XEL), the prob- lem of grounding mentions of entities in a for- eign language text into an English knowledge base such as Wikipedia, has seen a lot of re- search in recent years, with a range of promis- ing techniques. However, current techniques do not rise to the challenges introduced by text in low-resource languages (LRL) and, surpris- ingly, fail to generalize to text not taken from Wikipedia, on which they are usually trained. This paper provides a thorough analysis of low-resource XEL techniques, focusing on the key step of identifying candidate English Wikipedia titles that correspond to a given for- eign language mention. Our analysis indi- cates that current methods are limited by their reliance on Wikipedia’s interlanguage links and thus suffer when the foreign language’s Wikipedia is small. We conclude that the LRL setting requires the use of outside-Wikipedia cross-lingual resources and present a simple yet effective zero-shot XEL system, QuEL, that utilizes search engines query logs. With experiments on 25 languages, QuEL shows an average increase of 25% in gold candidate re- call and of 13% in end-to-end linking accu- racy over state-of-the-art baselines.1 1 Introduction Cross-lingual Entity Linking (XEL) aims at ground- ing mentions written in a foreign (source) language (SL) into entries in a (target) language Knowledge Base (KB), which we consider here as the English Wikipedia following Pan et al. (2017); Upadhyay et al. (2018a); Zhou et al. (2020). In Figure 1, for instance, an Odia (an Indo-Aryan language in India) mention (“Chilika Lake”) is linked to the corresponding English Wikipedia entry. The XEL ∗ Both authors contributed equally to this work. 1Code is available at: http://cogcomp.org/page/ publication_view/911. The LORELEI data will be available through LDC. XEL: https://en.wikipedia.org/wiki/Chilika_Lake Gloss: Flood appeared in Arakhkud near [Chilika Lake]. [ଚିଲିକାେର] Chilika Lake Odia: [ଚିଲିକାେର] ଅରଖକୁଦ ନିକଟେର ଏକ ମୁହାଣ େଖାଲିଥିଲା । English: Figure 1: The EXL task: in the given sentence we link “Chilika Lake” to its corresponding English Wikipedia task typically involves two main steps: (1) candi- date generation, retrieving a list of candidate KB entries for the mention, and (2) candidate ranking, selecting the most likely entry from the candidates. While XEL techniques have been studied heavily in recent years, many challenges remain in the LRL setting. Specifically, existing candidate generation methods perform well on Wikipedia-based dataset but fail to generalize beyond Wikipedia, to news and social media text. Error analysis on existing LRL XEL systems shows that the key obstacle is candidate generation. For example, 79.3%-89.1% of XEL errors in Odia can be attributed to the limi- tations of candidate generation. In this paper, we present a thorough analysis of the limitations of several leading candidate …

Content truncated to 500 words